The 2.8 Megawatt, 18,500 square foot retrofitted data centre is a Tier 2 facility, meaning it isn't as resilient as Tier 1 facilities. It's pretty efficient though, according to Cisco, which put its Power Usage Efficiency measure at 1.4, compared to 1.34 at the company's Allen, Texas, facility and 1.9 in Richardson, the two data centres the new RTP1 facility will initially back up.
Cisco plans to have RTP1 serve as a production backup for other Cisco global data centres, says company IT official Jim Cribari. It's also a showcase of Cisco technology and implementation techniques for customers, and to prove that they do not need a separate physical facility for DR.
"It shows customers how to maximize their investment in DR and application development environments," Cribari says.
RTP1 is the oldest building in Cisco's Research Triangle Park campus. As a data centre, it has a capacity of 438 racks housing up to 5,000 Unified Computing System (UCS) blades and 125,000 virtual machines, with an average usage of 6.5 kilowatts per rack. The facility is both air and chilled water cooled - "water-side economizers" allow the mechanical chillers to cycle off when air can be used to cool the facility.
RTP1 also has a hardened floor instead of a raised floor - all cabling and other conduits and services are on overhead rails.
The backup capability is based, not surprisingly, on Cisco's UCS and Nexus switches, and NetApp's SnapManager mirroring and replication software. Service profiles defined in UCS servers enable the system to effectively emulate the data centre workload, while virtual routing and forwarding (VRF) capabilities in the Nexus switches mimic the primary data centre's infrastructure.
Service profiles are defined, provisioned and enforceable attributes of an application or VM that follows that VM as it moves around an enterprise. VRF allows multiple instances of a routing table to co-exist within the same router at the same time. SnapManager performs backup, restore and disaster recovery operations for virtualized servers on a NetApp storage system.
Together, these three components allow RTP1 to serve as a disaster recovery site based on "virtual data centres" created within the physical data centre. It can serve as a DR site for multiple production data centres, Cisco says, by defining subsets of applications based on their importance to the enterprise.
RTP1 houses a 40-blade UCS cluster supporting 800 to 1,000 VMs, which in turn house service profiles defined in advance to back up a data centre site. All UCS chassis are connected by Cisco's Unified Fabric architecture to top-of-rack Nexus 6100 switches, which are then connected to Nexus 7000 switches in the network backbone. The Nexus switches also connect Fibre Channel traffic to the storage fabric.
The Texas data centres crunch 1,300 production applications, but not all can be backed up simultaneously; Cisco IT defines four categories of criticality for the applications, accommodating the most critical first and then the less critical, as capacity allows.